INDEX

Explanations

phrases related to implying or suggesting something, often involving controversial or speculative content

oai_token-act-pair · gpt-3.5-turbo

phrases that imply causation or suggestion of responsibility

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

SEA

-0.79

Ĥ¬

-0.78

chal

-0.72

 fixme

-0.72

¥µ

-0.72

zanne

-0.68

Lago

-0.68

externalActionCode

-0.67

hyde

-0.67

lda

-0.67

POSITIVE LOGITS

oded

1.31

osion

1.22

anting

1.21

icating

1.18

icate

1.15

icates

1.12

ausible

1.12

oding

1.11

anted

1.06

 impl

1.04

Activations Density 0.006%

phrases related to implying or suggesting something, often involving controversial or speculative content

phrases that imply causation or suggestion of responsibility

No Comments

No Known Activations